The watermarking based on wavelet fusion gains sharp edge and discontinuity in watermarked image due to embedding in smooth region. To solve this problem spatial fusion watermarking method is proposed in this paper. The image is decomposed into four levels by predicting the range of intensity. To improve the capacity of embedding, higher counted region is chosen for embedding which indirectly chooses sharp region. The Text or ROI data is watermarked in the chosen region by differentiating the pixel with one by the constraint of even or odd between the cover image and text or ROI. The decomposed images are composed by averaging the pixels of the regions greater than one. After extraction, the watermarked medical image is reconstructed to original medical image by reversible property. This system is evaluated with various metrics using standard medical images which show good quality and high imperceptibility with embedding capacity. © 2013 Springer.